LIVING STANDARDS SURVEYS IN DEVELOPING COUNTRIES
Ramesh Chander, Christiaan Grootaert and Graham Pyatt, World Bank
data analysis.
Fortunately, other initiatives
are under way which are complementary to LSMS.
Most notable is the UN National Household Survey
Capability Programme, a global program to help
countries develop their capability to conduct
household surveys.
Equally, LSMS may cooperate
with regional efforts such as the Inter-American
Household Survey Program.
Some limitations of LSMS must be pointed out.
On subject matter, the focus of attention will be
on data on income and expenditure, which are seen
as the core determinant of living standards.
Nevertheless, the study of incomes and expenditures will be approached in a broad framework
including other components as well.
With respect to country coverage, it must be
said that while we hope that our dissemination of
methodological findings will allow the largest
number of countries to enjoy the benefits of
LSMS, our direct involvement will be limited to a
handful of countries where we shall conduct
experimental survey work.
The projected time schedule of LSMS will place
constraints on methodology.
It excludes e.g. the
possibility of undertaking longitudinal studies.
In order, however, to fulfil the main purpose of
LSMS, viz. to monitor the impact of development
policy, an emphasis on changes over time is
clearly indispensable.
Accordingly, the main
thrust of LSMS will be towards the design of a
set of concepts, survey questionnaires and procedures for analysis which can be repeated, e.g.
at 5 year intervals, to make statements about
changes over time.
Finally, the emphasis in LSMS is on measuring
how different groups participate in the development process.
The emphasis will therefore not
be on intra-group inequality, or the place of
individuals in their group, but rather on intergroup mean differences, or the position of groups
in society.
I.
INTRODUCTION
The purpose of this paper is to introduce
the Living Standards Measurement Study (LSMS),
recently begun at the World Bank.
The initial
objective of this study is to identify the data
which can most usefully be collected and the
best ways of collecting it, in order to provide
a description of living standards and a basis
for analysis in support of policy formulation.
The LSMS has been conceived in response to
the widely recognized needs for a better quantitative basis for the design and monitoring of
development policy.
Given that such policy is
primarily concerned with securing and raising
the living standards of people, there is a
premium on facts that relate to living standards
in general, and more specifically to poverty,
basic needs, employment and income distribution.
While the international statistical system
has focused its efforts on gathering data on
production, accumulation, internal and external
capital accounts, etc., along with the compilation of socio-economic indicators, only
limited efforts have been made to put these
diverse data sets together in a consistent
framework which could foster a better understanding of the development process.
The LSMS will
attempt to present the two sets of data in one
consistent accounting framework which will highlight the flows of income from the productive
system towards the households and other institutions who provide productive inputs.
Our main
interest, of course, is with the household sector
which is seen as an aggregate of socio-economic
groups.
Ultimately, therefore, the LSMS wishes
to answer the question: which groups in society
benefit from economic development and which, if
any, are left out? With such knowledge, policies
can be formulated or reformulated to distribute
or redistribute the fruits of the development
effort.
Household surveys are seen as the major tool
to collect data on living standards.
The current
situation in most developing countries regarding
household surveys can be considered deficient in
a variety of ways.
The problems are well known:
inappropriately defined concepts (e.g., of the
household and income) often lead to internal
inconsistency of the data collected due to
varying enumerator interpretations.
Existing
sampling frames are often inappropriate for
household surveys and/or outdated.
Lack of
resources, especially trained manpower, plagues
the field operations and results in the collection of poor quality data.
Serious bottlenecks
exist at the edit and data processing stages,
in terms of software, hardware, and personnel.
Although not exhaustive, the above list of
issues that could potentially be addressed by
LSMS is sufficiently long to show a clear need
to define a relatively narrow focus at the outset.
The areas where this study can be of most
help are determined very much by the comparative
advantage that the World Bank has accumulated
through its research on poverty, income distribution, etc. The LSMS will therefore focus
on conceptualization, survey methodology and
II.
THE WORK PROGRAM IN OUTLINE
The work program presented below has benefitted
from the meeting held at the World Bank last
February of an Expert Advisory Group consisting
of leading economists, government statisticians
and representatives of international organizations.
It will evolve in three phases broadly corresponding to the sequence design, implementation,
and analysis - spread over four years.
This
order of succession, however, is not intended to
prevent overlap which, in fact, is needed to
ensure proper feedback between the phases.
The
feedback from analysis to design is considered
especially crucial.
Phase I includes an evaluation of existing data
and methods of data collection and analysis, as
well as the preliminary design of new surveys and
of a series of topic studies which are envisaged
as one of LSMS's main outputs.
Based on the
proposals generated in Phase I, Phase II will
carry out the testing of survey instruments and
the collection of data in a few selected countries. Phase III will cover the analysis of the
newly collected data, the write-up of the topic
studies, and the finalization of survey manuals
83
B.
PreliminaryDesign of Topic Studies
Work on the preliminary design of four of the
six proposed topic studies will commence this
summer, focusing on the specification of concepts,
pro forma tabulations, and the analyses which we
would like to carry out with Phase II results as
available.
The following notes are a preliminary
guide as to possible content of the studies.
Income Distribution and National Accounts:
This study will show how household survey data can
be used as a micro-data base for extending the
scope of conventional national accounts. The
emphasis will be on national accounts in the
expanded social accounting matrix (SAM) format in
which the household sector is subdivided into
socio-economic groups in a flexible manner.
Demand Analysis: This study will examine in
depth the various issues which underlie estimates
of total consumption by households.
The subject
matter will cover Engel curve analysis and household composition effects, together with problems
of price variations between households, seasonal
effects, and the need to impute values to income
in kind, production for self-consumption, and
various benefits obtained from the use of public
services. A second dimension of this study will
be the analysis of data on housing conditions and
the accumulation of durable goods, since such
analysis is likely to supplement our understanding
of consumption patterns and economic well-being.
Poverty and Inequality: This study will
address issues of poverty and inequality from the
perspective of societies stratified according to
socio-economic groups.
It will explore indicators
of poverty and affluence which could capture
elements of living standards that complement the
information in indices such as household per
capita consumption. Particular emphasis will be
given to identification of the poor and the characterization of poverty, particularly in terms of
identification of the factor endowments of households.
Manpower, Employment and Earn in~s : The study
will cover the sources of income for households of
different types with particular emphasis on
employment income and income from household production activities. The analysis of dualities in
labor markets will be a critical theme, based
largely on the estimation of earnings functions
which recognize education level as a key variable
along with other household resources, such as
access to land. Labor force participation and the
distinction between primary and secondary workers
will provide still another element of this study.
The last two topic studies are: Economic
Influences on Demography and Access to Public
Services. Although LSMS focuses on income and
expenditure, it is recognized that health,
nutrition, fertility and education are important
elements of living standards. These two studies
will attempt to explore these topics keeping in
mind that the usefulness of survey results in
these areas is greatly enhanced if they are compatible with corresponding economic indicators.
Accordingly, LSMS will extend itself into the area
of economic and demographic linkages because
(a) the demographic component that is required has
been well thought through and thoroughly tried by
the World Fertility Survey; and (b) these particular
types of linkage are already well researched,
particularly within the Bank.
and questionnaires on the basis of our experience.
In line with the need for constructive feedback
Phase 2, the field work stage, overlaps with the
other phases, such that some experimental fieldwork may take place before the completion of all
conceptualization and, equally, preliminary
analysis may begin before all the data are
collected.
The list of topic studies for LSMS is not
final. So far we have identified the following
six potential topics, to be discussed in further
detail in the next section:
i) Poverty and
inequality; 2) Manpower, employment and earnings;
3) Income distribution and national accounts;
4) Demand analysis; 5) Economic influences on
demography; 6) Access to public services.
The general form of each study will be:
(a) concepts to be measured and their theoretical
basis; (b) questionnaire modules to measure the
concepts; (c) numerical results, with special
emphasis on tabular and graphical presentation;
(d) derived statistics and their interpretation;
(e) more advanced analysis, hypothesis testing
and interpretation; (f) policy implications.
The primary output of LSMS will thus be a set
of studies on what data is needed and how it can
be used to illuminate particular areas of policy
intervention.
By giving this emphasis to
applications of household survey data and demonstrating their use~ we hope to contribute as much
to the stimulation of demand for better surveys
as to their design.
In order to keep the professional audience
informed of the progress and results of LSMS as
it proceeds, the study's findings will be disseminated via scientific journals and meetings
in the usual manner, as well as through a newly
established LSMS Working Paper Series to be
available on request. This series will begin
with the papers presented to the Expert Group
meeting in February.
III. PHASE I
A. Evaluation of Existing Data and Methods of
.Collection and Analysis
This activity has already begun, and papers
presented to the Expert Group meeting held in
February 1980, document much of the experience to
date. This documentation will now be extended by
further papers on specific investigations of
survey methodology which might be set up as new
experiments, dealing e.g. with alternative recall
periods, use of different survey methods to
improve data quality and with the reduction of
non-response.
While there is enormous scope for further
analysis of existing data, the general position is
that enough has been learned to date to suggest
that this will be a less productive activity than
working on fresh, newly designed information.
Accordingly, such analysis will be restricted to
situations which are exceptional either in terms
of the richness of the information source and
prospects for local collaboration, or because
such work will provide the additional opportunity
to become familiar with the situation in a country
which is likely to be involved in subsequent
stages of the LSMS.
So far, five countries have
been identified which are exceptional in these
terms.
84
the extent that an interface is required to relate
information on consumption to that which is
collected from production units; and where the
household is the production unit, there is
obviously a strong case for incorporating information on both its production activities and
consumption in the same survey.
Hence the household needs to be carefully defined both as a
producing and as a ~onsuming unit.
The above is most important in relation to
households engaged in subsistence agricultural
production.
But in fact most households provide
both goods and services for themselves, if only
at the level of growing a few vegetables and
maintaining property.
The less developed the
society, the more prominent are such activities;
and they are generally of greater relevance for
the poor as compared to the rich.
It is therefore
important to measure these activities in a systematic way, as a complement to the picture of
incomes and consumption which can be obtained
from details of cash flows and transactions.
Their valuation raises difficult questions of
imputation, which cannot always be neatly resolved
with reference to economic theory.
Essentially
there are two approaches, one based on the
estimation of what it would cost to buy the goods
or services which the household provides for
itself, the other on income forgone by the household in choosing to supply goods or services to
itself, rather than work for wages.
In an ideal
world these two approaches might yield essentially
the same answers.
But the real world is not so
simple and, in particular, the opportunities to
work remuneratively can be strictly limited.
Given these general considerations, some com ~
ment can be offered on the particular topics
which might be covered.
Enumeration of Household Members:
The starting
point is evidently with the members of the household, covering age, sex, race, education and interpersonal relationships.
Next, information needs
to be collected on the boundaries of the household as a producing unit, implying a designation
of the occupational status of each individual.
Employment and Occupation:
As a check on
individual occupational status, and because the
results could be of direct interest in their own
right, there is a case for including questions
on time use within LSMS. Not least, such data can
provide a means of checking on the mixture of
activities undertaken by individuals, and hence
on the items of income and expenditure which need
to be imputed.
They can also lead naturally into
details of employment status and sources of income.
The technical issues of collecting time use data
are considerable.
This then may be an area in
which LSMS should experiment.
H ousin ~ and Durables:
Data on housing should
include the availability of amenities such as
fresh water, waste disposal, and lighting, and
might potentially be extended to include household
durable goods such as cooking and heating
appliances.
Similarly, the section on acquisition
and sale of durable goods should cover other topics,
such as weddings and funerals, which give rise to
significant, irregular and infrequent expenditures.
This category must also include data on asset
accumulation and sales associated with household
production activities.
Cash Transactions:
Next, a full balanced
LSMS may also attempt to make use of community
or village level surveys.
These can contribute
substantially for three reasons:
(i) while the
emphasis we would like to put on changes in
living standards over time is hard to capture
simply through point-in-time household surveys,
it is much easier to track fixed communities
over time and, hence, conduct longitudinal
surveys of these.
(2) Some information concerning
the availability of services can perhaps more
usefully be collected by community level surveys.
(3) One of the major problems with household
surveys is that they can easily become too long
and/or unmanageable.
A potential advantage of
setting household and community surveys in a
common framework is, therefore, that this
provides some opportunity for relieving the
pressure on the household survey questionnaire
and making it simpler.
Our thoughts in this area
are, however, embryonic and call for considerable
development before it can be decided to what
extent LSMS will involve itself in data collection
at the community level rather than at the household level.
C. Preliminary Design of New Surveys
The specifics of survey content raise a host
of issues which must ultimately be resolved in
relation to the details of the topic studies.
Here we will make no attempt to be exhaustive,
but rather to comment on some specific points
which seem to be of strategic importance.
Sample Design:
First, the sample frame for
collecting data will necessarily be based in the
first instance on geographic subdivisions.
It is
therefore not essential to canvass nationwide
surveys in any year.
However, there would seem
to be a strong case for doing so, since the
geographic subdivision of households is only one
of several relevant groupings.
The value of
results will be much reduced if nationwide surveys
are eschewed.
Similarly, it would seem to be
essential to collect data over a full annual cycle,
if only because of the potentially important
seasonal effects which can be captured as a result.
Second, in designing samples, the emphasis
should be towards good estimates of mean values
for households of a given type, rather than on
variations within household types. As argued in
the introduction, the objective should be to
measure, first, average levels of living within
groups of households, and hence the changes in
them over time, and, secondly, the differences in
these averages which arise between groups.
The
efficiency of estimation of intragroup variations
is not crucial, and therefore is not an important
factor in sample design as far as LSMS is concerned.
Household as a Producing and Consuming Unit:
Since many households in developing countries are
engaged in household and/or unincorporated
business activities, it will be important to
define carefully the boundaries of the household
as an income earning/consuming unit.
Adopting a
narrow definition potentially simplifies the
household survey.
Equally, data which are not
picked up in one inquiry must be sought in another
to the extent that they are important.
It can be
argued that one of the key areas in which data are
needed is the relationship of the household to the
production structure of the economy.
Such data
must therefore be collected from households to
85
statement of cash incomes and expenditures is
called for, providing a consistency check on more
detailed information and, at the same time,
relating the household
to the cash economy.
In
collecting data on cash expenditures, an important point is whether quantity or price data
should be sought to supplement the information on
financial outlays.
A variant which may be of
interest would be not to attempt either for
individual households, but rather to seek price
data independently, e.g. by having separate
surveys of prices conducted among a selected group
of housewives who might be interviewed individually
or as a group, or alternatively by collecting
data from markets or retail outlets.
It can be
noted that this type of price data is also needed
in the International Comparisons Project work on
international purchasing parities and for generating national accounts at constant prices.
Furthermore, a vector of prices is always needed
to impute values to those items which are not
purchased for cash:
the approach offers one way
of obtaining these vectors.
And finally, surprising variations are observed in prices for
consumer goods both over time and according to
the haggles or bargaining which are a part of
everyday life.
This type of variation is a component of the intragroup variation which we have
previously suggested is not a primary concern.
Hence it may be unnecessary to pursue it.
On the income side, cash receipts derive from
employment, transfers, returns to financial assets
and sale of own produce.
Of these, it is reasonable to anticipate better data under some heads
than others.
With respect to cash wages, the
potentially important feature, beyond the amounts
involved, is to fix the nature of employment so
that the household can be linked up to the production structure.
This essentially involves a
description of the type of work done and the nature
of the employing agent or establishment.
One
possible way of fixing links to the formal sector
of production arises when the latter can be
identified through payment of social security contributions.
This is perhaps an area for experimentation.
Data for some types of transfers are easier to
collect than others.
Cash transfers between households will arise both on the receipts side and as
items of expenditure.
Hence they can usefully be
cross-checked at the aggregate level.
Transfer
receipts from social security, scholarships, and
remittances do not present any severe problems,
but the transfers which derive from corporate
distributions are much more difficult to pin down,
as are interest payments and receipts.
There is
little that can be done to improve the quality of
such data obtained at the household level of
inquiry.
Finally, sales of own produce are an important
element of the overall picture of cash transactions.
In subsistence societies such sales are likely to
be irregular and subject to a marked seasonality.
The reference period is therefore important.
Equally, the nature of the 'own product' sold is
important for relating the household to the production side of the economy.
Non-Cash Transactions:
Non-cash transactions
which do not add to the stock of assets are simultaneously an element both of incomes and expenditures.
They can be measured in physical terms
86
via inputs (notably labor time) or outputs (such
as kilos of vegetables).
But however this is
done, the same physical entry appears both as an
item of consumption and on the income side of the
accounts:
and it is a separate question to determine what value to associate with each physical
magnitude.
These non-cash elements of income and expenditure which need to be imputed fall under four main
headings.
The first is the amount to be imputed
as a result of ownership of durables, notably
housing.
The second is attributable to income in
kind, such as meals provided by the employer at
the place of work.
The third category covers the
goods and services which households provide for
themselves via direct labor input. And finally,
there are the services provided by the state which
households utilize.
This latter category includes
household benefits from state education and health
facilities, and hence is important in broadening
the picture of a household's living standard as
otherwise determined.
Difficult questions of
imputation arise in relation to such items; nevertheless, it would seem to be worth pursuing them
because of the obvious links with policy and
welfare.
The above discussion of survey content can be
summarized by suggesting that data might be
collected in six areas as follows:
(I) Enumeration of household members with
details according to age, sex, race, education
level, occupational status and interrelationships.
(2) Individual time use by locations, activities
in each location, and mode of transport between
locations in order to cover employment status and
other aspects of labor force participation.
(3) Characteristics of the dwelling, including
household durables, amenities, ownership, and rent
paid, if any.
(4) Infrequent expenditures, i.e. acquisition
of household and personal durables, home improvements, weddings, funeral expenses, etc., over the
past 12 months, plus acquisition and sale of
assets for use in household production activities.
(5) Full accounting of cash transactions:
expenditures by item, incomes by source.
(6) Consumption of own produce and public
services by type of goods or services, plus income
in kind.
All this adds up to being an ambitious undertaking.
The necessary details might well be cut
back by a determined attempt to restrict information to that which can be used for checking the
internal consistency of the data collected or to
that which is of direct relevance for predetermined
tabulations and analyses.
Equally, it may be
thought that some of the suggested topics could
well be deleted.
We fear, however, that the
pressures will be in the opposite direction,i.e.
to attempt more rather than less.
IV.
PHASE II
The work on Phase II, based on the proposals
generated in Phase I, will cover fieldwork in
selected countries, involving both small-scale
experiments or pilot surveys and, in some instances,
full national surveys.
In any one instance, the
detailed content of the survey work will be based
on a range of considerations, including the
particular interests of the country in question,
while staying within the overall conceptual frame-
variables.
Again, there seems to be no substitute
for direct recommendations in this area, and for
some worked-through examples which can provide a
ready checklist for others to follow up. The
third level of results comprises the graphs,
tabulations, and standard formats for presentation
(and publication) of what has been revealed by the
survey.
Some countries are well able to carry
through this stage of a survey without any such
guidance.
But there are a great many which are
not. A significant amount of the current frustration with the "state of the art" is due to the
fact that the processing of results at this stage
is ill-designed regarding both scope and content.
Moreover, it is the results obtained at this
level which are of the greatest potential interest
for public debate.
Comparing such results over
time will define the changes which have taken
place, and hence the evolution of living standards
which is our primary concern.
Therefore, such
results merit the closest consideration.
Finally, results take the form of in-depth
investigations, typically involving multivariate
analysis, hypothesis testing, etc. There is a
great shortage of individuals who can design and
execute studies at this level, e.g. studies on
Engel curves; hedonic studies of housing conditions;
measures of poverty and inequality; and earnings
functions.
Such studies can add depth to the
understanding obtained from simple, tabular results,
and hence be of value to the policy-making process.
Accordingly, they should be a part of the LSMS, in
the spirit that prototype studies might usefully
be replicated in countries where the necessary
skills are available.
work of LSMS.
As the topic studies are developed and their
data requirements are translated into draft
questionnaires, a number of technical, statistical
issues will come to the fore.
These include the
practical problems of measurability, required
accuracy and correction procedures.
There will,
therefore, be a need to develop a series of
survey documents or manuals on methods and procedures for data collection, covering questionnaire design, sampling and field methods, and
guides to editing, statistical imputation and
tabulations.
These should be based on experimental testing to the extent that resources will
allow, and result in the documentation of recommendations on how to implement the data collection with which LSMS is concerned.
The development of this aspect of LSMS will
depend on anticipated and revealed problems in
implementing LSMS data collection.
The core
questionnaire will not only need to be translated
and modified to suit local situations, but the
need to have alternative versions of the questionnaire for different groups (literate vs.
illiterate, those engaged in subsistence production, etc.) can be anticipated at the outset.
Similarly, there is known variation in field
methods, e.g. with respect to numbers of interviews, respondents and supervisors.
The identification of best practice over a
range of countries through the success actually
achieved and the obstacles encountered provides
an initial means of identifying the extent to
which modifications of the core questionnaire
will be necessary and alternative formats for
particular modules will be required to meet
different situations.
Beyond this a more detailed
work program will be prepared on the range of
statistical experimentation which could efficiently contribute to the technical aspects of LSMS
survey implementation.
Vl.
CONCLUSION
In supplying essential data for the determination of policy, the potential importance of LSMS
is unlikely to diminish in the years to come.
This is firstly because of the historical trend
in most cultures towards a greater concern for
equality as such.
This trend is encouraged by the
weight of evidence that aggregate growth is not
always shared equally within countries.
Furthermore, because many countries must now address the
probability of having to restructure their
economies in the face of slower overall growth of
export demand, and a worsening of the terms of
trade, there is a growing need to provide the
facts on how living standards are linked to the
production structure as well as how they are
changing over time. Without such information
there will be major difficulties in formulating a
policy for restructuring which is sensitive to its
repercussions on different socio-economic groups,
and in judging the success of any policy in protecting living standards.
When data are to be imputed in the policy making
process, their timeliness is of utter importance.
At present, many countries who are able to execute
surveys in the field, run into major problems at
the stages of editing, tabulating and analyzing
the data, and many surveys are therefore slow to
yield useful results.
This, in turn, is a significant reason for the low priority given to survey
work by many governments.
This means that opportunities for basing policy on sound and recent data
are being missed. If this is so, customer education
via demonstration effect should be a top priority
for LSMS, and processing and analysis of data to
V.
PHASE III
Phase 3 is the final phase of LSMS and covers
the analysis of the collected data, the documentation of the results, the recommendations on the
appropriate approach that countries ought to
take, and the dissemination of all this information.
It should be emphasized here that the
documentation will include both statistical
materials such as questionnaires, sampling and
field management manuals, guides to editing, etc.
as well as the topic studies.
The latter are
also manuals in the sense that they are not
intended to be advanced treatises but rather
guides on how to use results from household
surveys to inform a broad audience about living
standards and to analyze their determinants.
The analysis of the survey results can be considered on four levels at least.
First, there
are the crude numbers obtained from field work,
edited for quality control and made accessible in
data files.
This product is the primary goal,
since it provides a micro-data set which should
be accessible in response to whatever specific
questions might be addressed to it, or whatever
working files might be needed for particular
applications and analyses.
Second, results take
the form of derived statistics which can usefully
complement the basic data, such as subtotals,
ratios, and imputed values for particular
87
In view of the range of possibilities as to
how LSMS might influence future work on living
standard measurement, the study will not be
directed to any fixed ends beyond those described
in this work program.
By maintaining the close
cooperation there has been with the UN National
Household Survey Capability Programme through
the early stages of defining the LSMS and by
keeping the Statistical Commission informed of
progress on the study, it is hoped that the LSMS
research study will lead to a set of results
which might eventually be commended as guidelines
for national practice.
draw out its policy implications deserves the
greatest emphasis.
As a research study, the outcome of LSMS is to
some extent uncertain, and there will be a need
at the end of the study to undertake a major
evaluation.
Its initial success will depend on
the extent to which national statistical authorities choose to adopt the proposals which emerge
from Phase I of the study without necessarily
waiting for all phases to be completed.
Within
this, the LSMS recommendations on how to tabulate
and present household survey information may have
far-reaching and more immediate consequences
than full adoption of the LSMS proposals.
88